22 January 2024

Opinion: 'We need to invest in a national AI infrastructure now'

To avoid losing the battle for generative AI to dominant players such as America and China, the Netherlands must invest like a hare in an AI infrastructure. For our safety and economic security, argue concerned scientists from three universities.

Today, the Ministry of the Interior announced its long-awaited vision of the Dutch government on generative AI. To the general public, this technology is mainly known through the chatbot ChatGPT. But in the future, this technology could have a major impact on healthcare, energy transition and mobility issues, for example.

Consultancy firm McKinsey expects generative AI to increase productivity in all kinds of sectors: globally, $2,600-4,400 billion in value will be added annually, equivalent to Germany's GDP.

To secure its opportunities and interests, the Netherlands must retain and strengthen knowledge about AI. That knowledge is plentiful now. But developing, training and using generative AI models and retaining talent also requires a proper AI infrastructure: computers to understand, calculate, train and apply these models.

Immediately

That is why we as researchers at TU Delft, UvA and RUG and the Dutch AI Coalition and the national Innovation Centre for AI advocate building such a joint national AI infrastructure immediately. If we do not, we will remain dependent on the big US and Chinese tech companies that currently dominate the AI market.

A simple example of why this is necessary: with the energy transition, our electricity will increasingly be generated by solar panels and wind turbines with production varying greatly depending on weather conditions. At the same time, electricity grid capacity is limited. This complex puzzle must be solved to ensure that everyone can charge their cars and heat their homes. This requires AI systems. For such systems, essential for society, we should not depend on commercial platforms from the United States and China.

Values

We as the Netherlands, and as Europe, must also safeguard our values by focusing on human-centric AI, keeping an understanding of how personal data, copyrights, correctness of answers, bias and influence are handled. Big tech companies now dominate the debate around AI security with their own interpretations. It is a utopia to think we can realise our public values and safety with mere legislation and without understanding how AI systems are trained and applied.

Of course, we also want to retain and attract talent to the Netherlands: generative AI plays an important role in innovation and the future earning capacity of the Netherlands. Talent moves to where it can best flourish. Brookings research shows that commercial platforms have 29 times more computing power than university platforms. Without their own AI infrastructure, scientists have to resort to foreign platforms and Dutch companies fall behind when all the data, trained algorithms, and talent reside abroad.

We estimate an investment of around €400 million, for the construction and operation of a joint AI infrastructure. This will give the Netherlands an infrastructure that can compete with Big Tech platforms and become available for research and innovation in knowledge institutions and companies. Such an investment is too big and too risky for a single player. Without a substantial contribution from the Dutch government, that AI infrastructure will not come about in the Netherlands and we will remain dependent on Big Tech.

Vulnerable

The architecture of existing generative AI systems is centralised: all knowledge and control is concentrated in one organisation and in one physical location. This makes it highly vulnerable and an interesting target for malicious actors. Creating our own AI infrastructure also creates an EU backup network for the benefit of our national security and resilience.

Therefore, as knowledge institutions and industry, we call on the government not to passively watch as the competitive position of the Netherlands is eroded, but to actively invest in an AI infrastructure that allows the Netherlands to maintain and strengthen its leading position.

This article was published on the Volkskrant website (in Dutch).

Vergelijkbaar >

Similar news items

>View all news items >

May 29

Building responsibly on foundation models: practical guide by Utrecht University of Applied Sciences and RAAIT >

Researchers from RAAIT have published a practical guide for organisations aiming to develop AI applications using foundation models. The guide supports responsible decision-making.

SER: Put people first in the implementation of AI at work >

In a new advisory report, the Dutch Social and Economic Council (SER) calls for a people-centred approach to AI in the workplace, warning of risks to jobs and social cohesion.

🌞 Open Space: AI meets Science Communication – will you take the stage? >

Are you working on AI with impact? Wondering how to talk about it with the world? Join us on Thursday 4 July for an open space afternoon on AI & Science Communication, co-organized by Amsterdam AI and NEWS (the Dutch national centre for science & society).